为保证舰船模型三维重建的完整性以及细节程度，提出基于计算机视觉的舰船模型三维重建方法。基于双目立体视觉计算全局舰船目标点云坐标，生成舰船全局的三维点云数据；多视角点云配准方法对该数据进行配准后，输入多尺度特征递归卷积的稠密点云重建网络模型中，通过该模型生成舰船模型三维深度图，在此基础上，利用运动结构法完成舰船深度图中的三维曲线重建，对舰船模型进行颜色渲染，输出舰船模型三维重建结果。测试结果显示，该方法能够完成不同图像之间对应点的可靠匹配，确定各点的坐标位置；三维重建后模型的空间偏差均在0.015以下；能够较好地完成舰船结构的重建，重建后舰船模型的完整性较好，清晰呈现舰船的结构细节。
To ensure the integrity and level of detail of ship model 3D reconstruction, a computer vision based ship model 3D reconstruction method is proposed. This method is based on binocular stereo vision to calculate the coordinates of the global ship target point cloud and generate three-dimensional point cloud data of the ship as a whole; After registering the data using a multi view point cloud registration method, it is input into a dense point cloud reconstruction network model using multi-scale feature recursive convolution. Through this model, a deep 3D depth map of the ship model is generated. Based on this, the motion structure method is used to reconstruct the 3D curve in the ship depth map, render the color of the ship model, and output the 3D reconstruction result of the ship model. The test results show that this method can achieve reliable matching of corresponding points between different images and determine the coordinate positions of each point; The spatial deviation of the model after 3D reconstruction is all below 0.015; The reconstruction of the ship structure can be completed well, and the integrity of the reconstructed ship model is good, presenting the structural details of the ship clearly.
2024,46(6): 161-164 收稿日期：2023-06-14
DOI：10.3404/j.issn.1672-7649.2024.06.028
分类号：TP391
基金项目：2023年度河南省高校人文社会科学研究资助项目（一般项目）（2023-ZZJH-351）；黄河科技学院项目化课程教学改革项目（kg2022xm81）
作者简介：王玉(1986-),女,硕士,讲师,研究方向为三维动画。
参考文献：
[1] 肖世龙, 张德育, 张甜田. 基于虚拟现实技术的舰船航行三维图像重建[J]. 舰船科学技术, 2021, 43(2): 76-78.
XIAO Shi-long, ZHANG De-xue, ZHANG Tian-tian. 3D image reconstruction of ship navigation based on virtual reality technology[J]. Ship Science and Technology, 2021, 43(2): 76-78.
[2] 段渝, 杨青, 李中余, 等. 基于短时FrFT的双基雷达舰船成像最优时间段选取方法[J]. 现代雷达, 2023, 45(1): 18-25.
DUAN Yu, YANG Qing, LI Zhong-yu, et al. Optimal imaging time selection method based on short-time FrFT for bistatic radar ship target imaging[J]. Modern Radar, 2023, 45(1): 18-25.
[3] 张云, 化青龙, 姜义成, 等. 基于混合型复数域卷积神经网络的三维转动舰船目标识别[J]. 电子学报, 2022, 50(5): 1042-1049.
ZHANG Yun, HUA Qing-long, JIANG Yi-cheng, et al. Recognition of 3D rotating ship based on Mix-CV-CNN[J]. Acta Electronica Sinica, 2022, 50(5): 1042-1049.
[4] 柏正尧, 许祝, 张奕涵. 一种使用伪对应点生成的3D点云配准方法[J]. 计算机科学, 2023, 50(9): 210-219.
BAI Zheng-yao, XU Zhu, ZHANG Yi-han. Deep artificial correspondence generation for 3D point cloud registration[J]. Computer Science, 2023, 50(9): 210-219.
[5] 何鑫睿, 李秀梅, 孙军梅, 等. 基于改进Pix2Vox的单图像三维重建网络[J]. 计算机辅助设计与图形学学报, 2022, 34(3): 364-372.
HE Xin-rui, LI Xiu-mei, SUN Jun-mei, et al. Improved Pix2Vox based 3D reconstruction network from single image[J]. Journal of Computer-Aided Design & Computer Graphics, 2022, 34(3): 364-372.
[6] 白静, 孟庆亮, 徐昊, 等. ST-Rec3D: 基于结构和目标感知的三维重建[J]. 图学学报, 2022, 43(3): 469-477.
BAI Jing, MENG Qing-liang, XU Hao, et al. ST-Rec3D: a structure and target-aware 3D reconstruction[J]. Journal of Graphics, 2022, 43(3): 469-477.
